Setting up your first AI eval with a LLM-as-judge

I'm co-founder of Lovelaice and a product leader with 10+ years building products across fintech, payments, and compliance. I hold a CFA charter and have led AI product development in highly regulated environments — where AI failures aren't just embarrassing, they're liabilities.

I've watched smart teams make the same mistakes: choosing models based on benchmarks that don't reflect their use case, writing prompts that work in testing but fail in production, and leaving domain experts out of the loop. These aren't edge cases — they're why 80% of AI projects underperform.

Through these failures (my own included), I developed a systematic approach to AI experimentation that puts domain expertise at the center. I teach what I've learned building Lovelaice: how to test, evaluate, and iterate on A

I bring over 14 years of software development expertise and a decade of startup experience to help teams build AI products that actually work. After founding my first company six years ago, I run a consultancy specializing in helping startups build MVPs, solve complex technical challenges, and integrate AI effectively.

I've seen firsthand how AI projects fail due to lack of systematic experimentation—teams treat AI like traditional software and struggle with inconsistent results. That's why I co-created Lovelace, a platform designed for non-technical professionals to experiment with AI agents systematically.